| The rapid growth of visual media based applications necessitates sophisticated compression and indexing techniques in order to store, transmit and retrieve audio-visual information. The recent MPEG 4 and JPEG 2000 standards address the need for content based coding and manipulation of visual media. The upcoming MPEG 7 standard proposes content descriptors, which succinctly describe the visual content for the purpose of efficient retrieval. This implies that there is an impending need for efficient and effective joint compression and indexing approaches. Several compressed domain indexing techniques have been presented in the recent literature. These are based on the extraction of features from the compression parameters to derive indices for retrieval. However, there has been little work in the domain of exploring the use of these features to serve the purposes of both compression and indexing.; Humans are the predominantly the common users of visual data. The result of the revolutionary knowledge that researchers have gained in understanding the human brain in the last couple of decades, coupled with the promising performance of models that characterizes the human visual system, has motivated the design and development of image coders that correlate well with the human visual perception. However, there has not been adequate research directed toward visual data characterization based on human perception for the purpose of joint compression and classification.; In this thesis a joint compression and classification model for visual data in the wavelet domain is introduced, which operate at both the coefficient level (transform level) and the bit-stream level. While doing this, the importance of considering the perceived data by the human visual system in designing a joint compression and classification system will be evaluated. The concept of Joint Compression and Classification has been introduced, where the indices (classification parameters) are used to enhance the compression performance, rather than the compression parameters are only used as indices. This would result in a joint description of visual data that can serve for the purpose of both compression and classification. |
